AgentKit Engineer Kit Review: Are the 60+ Skills Worth It? (2026)

Prices and skill/workflow counts here are based on agentkit.best/engineer and the official docs, checked 2026-08-09. Numbers can change - always confirm on the product page before you buy.

Which AgentKit? Disambiguating four look-alikes

Read this before you scroll. If you searched "agentkit engineer kit," the results are a name-collision minefield. Four different products fight for that term:

• OpenAI AgentKit - OpenAI's Agent Builder / ChatKit / Connector Registry, launched 2025-10-06. A platform for building your own agents and chat apps. Completely different vendor and product.
• AgentsKit (agentskit.co) - a separate name-twin product ("89 agents, 103 skills"). Not this.
• ClaudeKit "EngineerKit" (docs.claudekit.cc) - docs for a similarly named kit. Also not this.
• This one: AgentKit for Claude Code at agentkit.best, driven by the ak CLI, formerly "Claude Kit." That's what I review and link to here.

If you landed here looking for OpenAI's builder or AgentsKit, you're on the wrong review - no hard feelings. If you want the ground-level context first, read what Claude Code is, then my full AgentKit review (the parent of this article) for the whole platform picture.

What is the AgentKit Engineer Kit?

AgentKit is a platform that turns Claude Code (and Codex or Copilot) into a working "AI dev & marketing team." It splits into two kits: the Engineer Kit (for coding) and the Marketing Kit (lead gen, email, SEO, CRM). The Engineer Kit is the developer-facing half, and the one most people buy.

It came out of the Claude Kit → AgentKit rename - the old ck CLI became ak, and the docs moved from docs.claudekit.cc to docs.agentkit.best. If you're migrating from the old tooling, I wrote a separate guide on going from Claude Kit to AgentKit (ck→ak). Worth flagging: this rename is unrelated to Anthropic's own @anthropic-ai/claude-code → @anthropic-ai/claude-agent-sdk change, which is a different thing entirely - don't let the similar timing confuse you.

Mechanically, the Engineer Kit is a layer on top of Claude Code, not a replacement for it. You still run Claude Code as your coding agent; the kit adds specialized subagents, skills, workflows, and slash commands that Claude Code can invoke. So the honest framing is: the kit is only as useful as your underlying Claude Code habit. If you barely use Claude Code today, the Engineer Kit won't magically change that - it rewards people who are already in the tool daily and want to level up the process around it.

The trickiest part is the numbers. The AgentKit homepage markets the whole platform as 108+ skills and 45 AI agents (17 Engineer + 28 Marketing). But if you buy only the Engineer Kit, what you actually get is 17 agents, 60+ skills, and 30+ workflows focused on engineering. Don't confuse "platform totals" with "what's in the kit you paid for" - the sales page sometimes blends them, and the /engineer page was even advertising "70+ skills, 40+ workflows" the day I checked . I use the conservative locked figures (60+/30+) throughout; if the live counts are higher, that's upside for you, not a loss.

What's inside: 60+ skills, 30+ workflows, 17 agents

This is the part neither the sales page nor the docs summarize for you. Here's how I'd group the Engineer Kit's 60+ skills into six buckets (roughly how the docs categorize them), with a one-line "what it's for" so you can picture it fast. Counts are approximate - treat them as proportions, not gospel .

Notice the heaviest bucket isn't "frontend/backend code" - it's Process & Methodology (~22 skills). In other words, the Engineer Kit sells you a workflow more than it sells you snippets. It nudges Claude Code into a brainstorm → plan → implement → review → ship rhythm instead of answering one prompt at a time.

On the agent side, the Engineer Kit ships exactly 17 agents (a heading sometimes says "15," but the list adds up to 17): planner, researcher, ui-ux-designer, database-admin, tester, code-reviewer, debugger, docs-manager, git-manager, project-manager, copywriter, journal-writer, brainstormer, scout, scout-external, code-simplifier, and fullstack-developer.

Then there are the workflows (30+): they orchestrate agents in sequence or parallel, ship prebuilt CLAUDE.md templates, and provide implementation-plan scaffolds. A typical workflow chains itself - scout maps the codebase → planner breaks the work into phases → fullstack-developer implements → tester writes tests → code-reviewer checks the diff, stopping only at the points where you need to approve. This is what keeps "17 agents" from being 17 disconnected tools: they get wired into pipelines that run end to end instead of being called one by one. Honestly, this is the part I rate higher than the raw skill list - you're buying a way of working, not just individual tools.

Not all 17 agents earn equal keep. After using it, here's how I'd rank them by how often I actually reach for them - no sugar-coating.

Worth it (near-daily):

• fullstack-developer - the workhorse that carries most end-to-end feature building.
• code-reviewer - reads a real diff, catches bugs/regressions before commit; nearly pays for the ticket on its own.
• planner - turns a vague request into clear phases; great for big tasks.
• debugger - traces root cause instead of patching symptoms.
• tester - writes and runs tests scoped to your change.

Useful in context:

• scout / scout-external - map a codebase and pull external docs; gold when you drop into an unfamiliar repo.
• researcher, database-admin, docs-manager, git-manager - strong exactly when you need them (schema, docs, clean git), quiet otherwise.
• ui-ux-designer, brainstormer, code-simplifier, project-manager - nice-to-haves, not must-haves.

Rarely touched by a pure dev:

• copywriter - leans content/marketing; backend devs barely touch it.
• journal-writer - technical journaling; lovely if you like keeping a record, skipped by many.

The straight take: the real value clusters in about five core agents plus the Process/Methodology group. If you expect to use all 17 every day, you'll be disappointed - and that's exactly what a 4.9/5 sales page won't tell you.

How much does the Engineer Kit cost?

This is where buyers get confused, because three separate things live on the same site: the kit, the bundle, and the Desktop app (a yearly subscription that is not required to run any kit).

Two things to keep straight. First, the $99 kit price and the $19/year Desktop subscription are different products - agentkit.best lists no recurring fee for the kit itself, and the kit runs fine with just the ak CLI. Don't buy the Desktop app thinking you need it to use the kit; it's a convenience layer, not a requirement. Second, the kit comes with lifetime updates and a money-back guarantee - the page states the guarantee but not a fixed duration, so read the refund terms at checkout . License keys are issued instantly.

The math is simple: if you only code, the $99 Engineer Kit is enough. If you (or your team) also do lead gen, email, SEO, or CRM work, the $149 Bundle is the better deal - $50 more for the entire Marketing Kit. For a deeper cost breakdown, see my dedicated write-up on AgentKit pricing - is it worth it?

Short version for the impatient: code-only → Engineer Kit $99; need marketing too → Bundle $149. Check Engineer Kit pricing — 20% off, now $79.20.

Installing and first run (ak CLI)

Setup was quicker than I expected - under five minutes. The ak installer is a native binary for macOS, Linux, and Windows: it auto-detects your OS/arch, downloads the stable build, verifies the SHA-256 checksum, and installs into ~/.local/bin. No Node or Bun involved. You authenticate with your email or an API key.

After you activate your license, confirm the agents and skills are present:

ak agents list
ak skills list --kit engineer

Inside a Claude Code session, the agents and skills then show up as slash commands and skills you can call directly - for example /ak:review to invoke the code-reviewer, or /ak:plan to kick off the planner. That's the whole loop: install, activate, list, use. If you're coming from the old ck tooling, note that a one-time ak migrate --from=ck handles the switch (it previews changes by default before you apply them). I keep the full step-by-step install, auth edge cases, and migration details in separate guides so this review stays focused on the buy decision.

Hands-on: how the Engineer Kit changed my workflow

Numbers on a page don't tell you whether a kit earns its keep in real work, so here's an honest before/after on a task I do by hand anyway: reviewing a pull request.

My test was letting code-reviewer go over a real PR. In the Claude Code session it's a one-liner:

/ak:review # run code-reviewer on the current diff

What worked: the agent anchored to the actual diff instead of handing me generic advice, ranked findings by severity, and pointed at specific lines. It flagged an off-by-one boundary comparison I'd skimmed past. For building features, fullstack-developer follows the plan → implement → test pipeline, so it wanders off-task less than I do when I prompt by hand. The payoff is repeatability: the same kind of task produces the same quality, regardless of whether I wrote a clever prompt that day.

What didn't: the agent occasionally reported duplicate findings, and you still have to read them yourself to filter noise - it does not replace human review. There's also a real learning cost: for the first week I kept the docs open to remember which agent does what. That friction is genuine and worth factoring into your decision.

One expectation to set: agents perform best when the repo already has a CLAUDE.md describing conventions. Drop them into an old, undocumented project with no conventions and the output gets diluted - you'll spend nearly as long fixing it as doing the work yourself. Put another way, the Engineer Kit amplifies a good process; it doesn't rescue a chaotic codebase. On projects I'd set up properly it felt like a mid-level dev running alongside me. On messy legacy code it was closer to an assistant that needs hand-holding.

The other shift is smaller but real: the prebuilt CLAUDE.md templates meant Claude Code "remembered" project conventions from the start of a session instead of me re-explaining them every time. That's the kind of quiet win that's hard to feel in a demo but adds up over a week of real work. The flip side - and I want to be fair about it - is that a chunk of the Engineer Kit overlaps with things a disciplined Claude Code user already does by hand with custom slash commands and a good CLAUDE.md. If you've already built that muscle, the kit's value is more "consolidation and consistency" than "brand-new capability." If you haven't, it hands you a battle-tested version of a setup that would otherwise take weeks to assemble.

Pros and cons

On balance it tilts toward "worth it" for the right person: the biggest downsides are the price and the early learning curve, not the quality of the agents. That's exactly why I score it 4.3 and not 5. The Engineer tier is also engineer-only - it doesn't include the marketing agents, which is fine if you just code but worth knowing before you buy.

Is the Engineer Kit worth it? Who should (and shouldn't) buy

Solo fullstack dev who uses Claude Code daily: yes. Getting a ready-made team for review/plan/debug/test beats building your own prompts and slash commands from scratch, and $99 with lifetime updates is reasonable - code-reviewer plus the Process group recover most of the value alone.

Team lead / agency: lean Bundle ($149) if any of the work touches marketing or content - $50 more is cheaper than buying the Marketing Kit later. The repeatability also matters more across a team, where consistent output quality is worth paying for.

Hobbyist or newcomer: hold off. Read what Claude Code is and run the plain tool first; you'll get more from the kit once you know what you're extending. There's no rush, and you can always buy later - the lifetime-updates model means waiting doesn't cost you a future upgrade.

Who should NOT buy: if you only occasionally dip into Claude Code, if your work is mostly outside coding, or if you're the type who already maintains a rich personal set of slash commands and a tuned CLAUDE.md, the marginal gain is small. Same if $99 in USD is a stretch relative to how much you'll actually use it - a kit you open twice isn't a bargain at any price. Being honest here matters more to me than the sale: the Engineer Kit converts best for people who were already going to spend real hours in Claude Code this month.

If you want to see how it stacks up against other options before committing, I compare it in AgentKit vs alternatives, and it also appears in my roundup of the best Claude Code kits of 2026. Already sure you're in the "should buy" camp? Get the AgentKit Engineer Kit — 20% off, now $79.20.

Frequently asked questions

Is $99 one-time or yearly?

The kit costs $99 and agentkit.best lists no recurring fee for it - you also get lifetime updates. Only the separate Desktop app is a yearly subscription, and it isn't required to use the kit. Confirm the exact terms at checkout before you pay.

How is this different from OpenAI AgentKit or AgentsKit?

Different products entirely. This Engineer Kit is part of AgentKit (agentkit.best) and runs on Claude Code via the ak CLI. OpenAI AgentKit is OpenAI's platform for building agents/chat apps (Agent Builder/ChatKit). AgentsKit (agentskit.co) is yet another name-twin. Same-sounding names, different vendors.

Is there a refund?

Yes, the site states a money-back guarantee, but it does not publish a fixed duration or conditions, so read the refund terms at the point of purchase before you buy.

Do I need Claude Code Pro?

You need Claude Code, and heavy agent use burns through usage fast, so a paid plan (Pro at $20/month or a Max tier) helps in practice. The kit itself doesn't force a specific Claude plan, but the free tier's limits will constrain how much you can run.

Does it work with Codex or Cursor?

AgentKit positions itself as supporting Claude Code, Codex, and Copilot. The smoothest experience is on Claude Code, which it's designed around; compatibility with other tools may vary .

Where do I buy it?

On the official site, agentkit.best (my link to it is in the verdict box below). Avoid third-party resellers - license keys are issued instantly on the official checkout.

Final verdict and where to buy

The Engineer Kit isn't magic, but it's one of the few bundles that turns Claude Code into a genuinely disciplined dev process - strongest at code-reviewer, planner, fullstack-developer, and the Process group. For developers who live in Claude Code, $99 with lifetime updates is a good deal; if you're new, learn the base tool first. My rating: 4.3/5.